August 2023 Summaries

The company had been working on an integrated prototype called VFNLU, which was no longer an academic exercise but a feasible product. After refining the model, they refactored their code to fit existing systems and added multilingual support. However, they soon realized that their ML platform's latency was too high, leading to a 3-month code refactor to redesign how their NLU dealt with requests. The company then started beta testing the VFNLU in April 2023, rolling it out to free users and finding bugs and deficiencies along the way. After 18 months of work, they released the VFNLU, which outperformed popular NLUs on benchmarks, supporting 30+ languages and having low training time. The company aimed to allow people to experiment with both NLUs and LLMs in the conversational AI world.

Developing a Center of Excellence (CoE) in conversation design (CxD) requires solidifying existing good processes, refining those in flux, and adopting strategies such as learning from others, defining clear roles and expectations, choosing the right tools, and creating a living style guide. By doing so, CxD teams can achieve repeatable success and maintain brand consistency. Building internal relationships is crucial for creating a shared library of resources and documenting successes. The pursuit of excellence in conversation design involves continuous refinement and sharing knowledge with peers to stay ahead of the evolving industry.

Voiceflow is a leading conversation design platform that helps customers design and prototype AI agents and hand off finalized designs to development teams for implementation in conversational AI platforms. The company has been quietly launching its end-to-end conversational AI platform features over the past six months, including LLM-powered knowledge bases, agent publishing and versioning, dialog management APIs, analytics, conversation transcripts, and customizable web chat widgets. Voiceflow aims to invest further in its capabilities to take on legacy platforms such as Dialogflow, IBM Watson, AWS Lex, and others, with a goal of enabling customers to collaboratively design, build, launch, and scale AI agents for various applications like customer support and in-app assistants.

SB is sharing their career journey as a conversational AI professional, starting with software engineering internships and eventually finding success in conversation design at companies like Rocket Mortgage and Instacart. They've learned the importance of explaining the value of their work to leadership, focusing on impact and data, and have developed best practices such as designing for users, listening to assistant speech, being mindful of space limitations, prioritizing accessibility, and ensuring inclusive language. SB emphasizes the need to define success criteria before starting a project, view failure as an opportunity for growth, and prioritize iteration and improvement over perfection. They also highlight the importance of advocating for oneself and educating others on conversational AI, particularly in industries where it may not be well understood.